No specific rehabilitative strategy has been shown to enhance lower extremity motor function and improve locomotion after a spinal cord injury (SCI) in man. Experiments on adult cats after thoracic spinal cord transection show that the lumbosacral cord isolated from supraspinal input can activate rhythmic stepping while the hindlimbs are fully weight supporting. In addition, this locomotor ability is strongly influenced by the type of motor training that the animal receives. Preliminary studies from two laboratories raise the possibility that paraplegic patients can be trained to generate a stepping pattern. In the 4-year randomized, prospective, controlled clinical trial, we will study the motor output from subjects with chronic thoracic SCI. A stepping motor output will be elicited using a partially assisted body weight support along with rhythmic oscillation of the lower extremities. Stepping will be facilitated by treadmill training that is similar to the procedure which has been used in the Program Project's studies with spinally transected cats. We will define the locomotor characteristics, as well as measures related to functional recovery and quality of life, that arise from this rehabilitative therapy. Evidence of extensor and flexor muscle activation and the timing of these activities during treadmill stepping will be determined from electromyographic (EMG) recordings in patients who range from complete sensorimotor loss below their SCI (graded Frankel A) to less severe impairments in subjects who retain some supraspinal descending influences on the lumbosacral segments and may be able to execute locomotion (graded Frankel B, C and D). About 26 SCI time from onset of injury, and residual strength (Motor Index Score). The study will include at least 3 matched pairs for each Frankel grade. One of the two subjects of each pair will be randomly assigned to body weight support and treadmill (BWS & T) training and will begin a 3-month course. When that subject completes the course, the second subject of each pair will begin. All subjects will initiate their training at 18 m/min with the minimum amount of weight support assistance necessary for them to step. Training will continue at that speed until they achieve an effective gait at full weight bearing, at which time they will advance to the next higher treadmill velocity. This training will be repeated until the maximum speed is achieved. We will evaluate effectiveness of the this intervention with repeated measures of a combination of physiological and biomechanical data obtained as subjects step on the treadmill and, if capable, walk overground, as well as by physiological and morphologic analysis of muscle function. These studies will also provide insight into some of the physiological actions of antispasticity medications, which will be used where appropriate after a subject's maximal treadmill velocity is achieved. We propose to use a highly stratified design with controls to provide scientific validity and statistical efficiency.